A COMPARATIVE STUDY OF ELASTIC AND NONLINEAR SOIL

RESPONSE ANALYSIS
Anna KARATZETZOU1,Stavroula FOTOPOULOU2, Evi RIGA3, Sotiria KARAPETROU4,
Grigoris TSINIDIS5, Evangelia GARINI6, Kyriazis PITILAKIS7, Nikos GEROLYMOS8 and
George GAZETAS9
ABSTRACT
Earthquake case histories of soil response over the last 70 years reveal the strongly non-linear,
hysteretic behaviour of soils. Over the years, numerous efforts have been made to model the nonlinear behaviour of soils during earthquakes. The paper presents a comprehensive comparative study,
performed in the framework of PRENOLIN project (https://www-cashima.cea.fr/), of the elastic and
non-linear response of two idealised soil profiles: a single layer soil deposit of 20 m depth with
constant shear wave velocity and an soil profile of 100 m depth with a gradient (parabolic) distribution
of shear wave velocity with depth. The profiles are subjected to a number of idealised and recorded
time- history accelerations.
The numerical simulations are performed with the research codes NL-DYAS, DEEPSOIL and
OPENSEES as well as the commercially available finite element programme ABAQUS, incorporating
several appropriate constitutive models. Results are presented in terms of acceleration time histories,
shear stresses and strains. The elastic and non-linear response computed with the aforementioned four
codes is compared with each other. Similarities and differences between the programs in predicting the
seismic response are highlighted.

INTRODUCTION
One of the crucial issues in earthquake engineering practice is the proper evaluation of the ground
motion. In the simplest case, this evaluation is performed through 1D site response analysis,
implementing the equivalent linear approximation (Schnabel et al., 1972) to model the soil dynamic
response. Although the method is cost-effective compared to more detailed non-linear dynamic
analysis, it does not account for several characteristics of ground motion in case of strong earthquakes
including the computation of permanent ground deformations. Actually, non-linear soil behaviour is
observed in all recent large earthquakes worldwide affecting seriously the site-specific ground motion.
1

Ph-D Candidate, AUTH, Thessaloniki, Greece, akaratze@civil.auth.gr

Post-Doctoral Researcher, AUTH, Thessaloniki, Greece, sfotopou@civil.auth.gr
3
Ph-D Candidate, AUTH, Thessaloniki, Greece, eviriga@civil.auth.gr
4
Ph-D Candidate, AUTH, Thessaloniki, Greece, gkarapet@civil.auth.gr
5
Ph-D Candidate, AUTH, Thessaloniki, Greece, gtsinidi@civil.auth.gr
6
Post-Doctoral Researcher, NTUA, Athens, Greece, geocvemp@yahoo.gr
7
Professor, AUTH, Thessaloniki, Greece, pitilakis@civil.auth.gr
8
Assistant Professor, NTUA, Athens, Greece, gerolymos@gmail.com
9
Professor, NTUA, Athens, Greece, gazetas@ath.forthnet.gr
2

It is therefore important to use adequate numerical tools that can describe efficiently and realistically
the soil non-linearity when predicting site response.
Several commercial or research numerical codes which are currently available, are capable of
performing non-linear soil response analysis. The analysis is commonly performed in the time domain
with the soil non-linear response being modelled with a non-linear constitutive model of various
complexity, ranging from a simple elastic-perfectly plastic model (e.g. a Mohr-Coulomb model) to a
more complicated and rigorous model (e.g. models that account for large strains, liquefaction etc). In
this paper we discuss representative results from a series of numerical analyses with four different
numerical codes on two idealized soil profiles. The objective of the analyses, which were performed
assuming either visco-elastic or elasto-plastic soil response is to cross compare the results of each code
in terms of accelerations, strains and stresses and through this comparison to qualitatively evaluate the
accuracy of each model. To extend our comparisons, the analyses are performed under several
assumptions regarding soil stratigraphy and properties, input motion characteristics, and the soilbedrock interface (elastic and rigid bedrock). The results are discussed in terms of acceleration time
histories, transfer functions, distributions of peak horizontal acceleration and shear strain and stressstrain loops.
DESCRIPTION OF CASE STUDIES
A series of elastic, visco-elastic and non-linear analyses are performed using two "canonical" soil
profiles P1, and P2, which represent ideal site conditions (Fig. 1). Soil profile P1 (Fig.1a) represents a
homogeneous single layer soil deposit of 20m in depth with constant shear wave velocity Vs=300m/s.
The shear wave velocity of the underlying bedrock Vs,rock is equal to 1000m/s. Soil profile P2 (Fig.1b)
resembles an inhomogeneous soil profile of 100m in depth with gradient shear wave velocity Vs,
varying with depth z according to the following generic form:

Vs Vs1 (Vs 2 Vs1 ) [( z z1 ) / ( z2 z1 )]a

(1)

where z1:depth at top of deposit (=0m), z2:depth at bottom of deposit (=100m), Vs1: Vs at top of
deposit (=150m/s), Vs2: Vs at bottom of deposit (=500m/s), : a constant (=0.25). The shear wave
velocity of the underlying bedrock Vs,rock is equal to 2000m/s.

(a)

(b)
Figure 1. (a) Profile P1, (b) Profile P2
2

A. Karatzetzou, S. Fotopoulou, E. Riga, S. Karapetrou, G. Tsinidis, E. Garini, K. Pitilakis, N. Gerolymos and G.

Gazetas
3

Compression wave velocity Vp is calculated as follows:

1
V p 2Vs 2

2 1

(2)

where v is the Poisson coefficient, assumed to be equal to 0.4 for sediments and 0.3 for rock. Mass
density is assumed to be equal to 2000kg/m3 for sediments and 2500 kg/m3 for rock.
Attenuation for elastic and visco-elastic analyses is characterized by a quality factor Q, which is
considered as follows:
(i) elastic analyses: Q=5000 for both sediments and rock
(ii) visco-elastic analyses: Qsed=Vs/10 for sediments and Qrock=200 for rock.
The corresponding attenuation is described in terms of viscous damping that is modeled in the
frequency depended Rayleigh type in ABAQUS and OPENSEES, while in DEEPSOIL and NLDYAS is assumed constant.
The elastic properties of the profiles are summarized in Table 1.In all cases, the analyses are
performed assuming dry conditions. To this end, the analyses are performed in total stresses.
Table 1. Elastic properties of the soil profiles
Soil profile z (m)
P1
0-20
bedrock
P2
0-100
bedrock

Vs (m/s)
300
1000
150-500
2000

(kg/m3)
2000
2500
2000
2500

Vp (m/s)
700
1900
360-1220
3700

Q elastic
5000
5000
5000
5000

Q viscoleastic
30
200
15-50
200

The degradation of the soil stiffness and the damping increase with increased shear strains are
described with proper G--D curves. More specifically, shear modulus reduction G/Gmax and damping
ratio curves are computed by equations (3) and (4):

G / Gmax 1/ (1 / ref )

(3)

min (max min ) ( / ref ) / (1 / ref )

(4)

ref max / Gmax

(5)

max sin
' v' (1 2K0 ) / 3

(6)

g z

(8)

where

'

'
v

g 9.81 m/s

K0 (1 sin ) OCR

30

OCR 1

Gmax Vs2

min 1/ 2Qs
max 0.25

(7)
(9)

sin

(10)
(11)
(12)
(13)
(14)
(15)

The resulting shear modulus reduction (G/Gmax) and damping ratio ( %) curves for soil profile P1 are
illustrated in Fig.2a. For soil profile P2, five sets of shear modulus reduction and damping curves are
calculated for depth intervals of 20m, referenced at 10, 30, 50, 70, and 90m depth (Fig.2b). In addition
to the hysteretic damping introduced by each non-linear constitutive model, a small amount of viscous
damping is assigned to account for the energy dissipation during the elastic part of the cyclic response.

(a)

(b)

Figure 2. Soil reduction G/Gmax and damping (%) curves for soil profiles (a) P1 and (b) P2.

Analyses are performed assuming constant shear strength for depth intervals of 20m. This
strength is estimated at the middle of each soil layer assuming a friction angle = 30. Considering the
increase of the mean effective stress with depth, soil shear strength increases with depth.
The analyses are carried out using both an impulse and a real accelerogram as input motion at
the model base. For the elastic and visco-elastic analyses we used two input motions: a Ricker type
pulse with corner frequency Tc=4Hz (Fig.3a) and the E-W component of a real accelerogram recorded
at station IWTH17 (Vs30= 1200m/s) of the Kik-Net array (Fig.3b). The non-linear analyses were
performed with the same accelerogram used for the elastic analyses, scaled at Peak Ground
Acceleration PGA equal to 0.5, 2.0 and 5.0 m/s2.

(a)

(b)

Figure 3. Input motions used for the elastic and visco-elastic analyses: (a) Ricker pulse, (b) real accelerogram

A. Karatzetzou, S. Fotopoulou, E. Riga, S. Karapetrou, G. Tsinidis, E. Garini, K. Pitilakis, N. Gerolymos and G.

Gazetas
5

In order to study the effectiveness of numerical codes to model either rigid or flexible
substratum, the analyses are performed for different boundary conditions at the sediment/bedrock
interface.
In the ensuing, results will be illustrated in figures with the following code name:
P[profile_no]_[damping]_[input]_Z[depth]_[bedrock], where:
profile no: can be 1 or 2 depending on the soil profile number
damping: can be el or ve for elastic and visco-elastic case respectively
input: can be puls or real for pulse like signal and for real accelerogram respectively
depth: can take values 0,1,2..10 for the accelerations and 0,1,...9 for the stresses and strains. When the
depth value is equal to zero, the acceleration time history is at free field conditions, while the stressstrain values are at depth equal to H/20, where H the soil profile depth.
bedrock: can be E or R for elastic halfspace and rigid bedrock respectively.

GROUND RESPONSE MODELING

Elastic, visco-elastic and non-linear ground response analyses are examined utilizing different soil
models. The important issues involved in the modelling of a soil profile subjected to earthquake
loading when the former behaves in the elastic, visco-elastic or non-linear range, are highlighted using
four different commercial and research numerical codes, namely: ABAQUS (ABAQUS, 2012),
OPENSEES (Mazzoni et al., 2009), DEEPSOIL (Hashash and Park 2001, 2002; Park and Hashash
2004; www.uiuc.edu/~deepsoil) and NL-DYAS (Gerolymos and Gazetas, 2005, Drosos et al., 2012).
ABAQUS and OPENSEES are generic finite element codes, while DEEPSOIL and NL-DYAS are 1D
soil response analysis programs. In all cases, soil deposits are discretized to account for the efficient
reproduction of the waveforms of the whole frequency range under study.
In ABAQUS and OPENSEES, the rigid base is modeled in simplified way by constraining
vertical displacements, while elastic base is modeled using the Lysmer and Kuhlemeyer (1969)
scheme, introducing proper dashpots. Time discretization is an important element for the analysis
accuracy. All codes offer an automatic time incrementation scheme, according to which the time step
of the analysis is properly selected and changed to achieve a stable solution. All codes will be
examined for the simplest case of 1D shaking, even if some of these codes (e.g. OPENSEES) are
capable of performing multi-dimensional earthquake loading ground response analyses.
A kinematic hardening model combined with a Von-Mises failure criterion and an associated
plastic flow rule is used to model soil non-linear response under ground shaking in ABAQUS
(Anastasopoulos et al, 2011). The model parameters are calibrated based on the given G--D curves
and the strength of the soil stratum. The analyses are performed with an implicit algorithm scheme,
using full Rayleigh viscous damping formulation and Masing rules to calculate the hysteretic damping
according to Masing unloading/reloading criteria. Full Rayleigh viscous damping as well as Masing
criteria is also used in OPENSEES. In OPENSEES we utilized the multi-yield surface plasticity model
with an associative flow rule as described in Ragheb, 1994, Parra, 1996, Yang, 2000. In DEEPSOIL
we used the embedded MRDF pressure-dependent hyperbolic model with non-Masing criteria, which
introduces a reduction factor into the hyperbolic model by Hashash and Park (2001), also available in
the code. DEEPSOIL provides a frequency independent damping formulation, as well as three types of
Rayleigh damping (simplified, full and extended, with one, two and four modes needed to define
viscous damping respectively). The recommended frequency independent damping formulation was
used for the analyses. NL-DYAS uses the smooth hysteresis model originally proposed by Bouc 1971
and Wen 1976, which was later extended by Gerolymos and Gazetas (2005), Drosos et al. (2012).
All models were properly calibrated using the aforementioned G--D curves. More specifically,
the targeted G--D curves (Fig. 2) were fitted either automatically (e.g. OPENSEES, DEEPSOIL) or
manually. In ABAQUS this was achieved by comparing the numerically estimated G--D curves,
computed by simulations of cyclic simple shear tests for different levels of shear strains
(Anastasopoulos et al., 2011) with the target curves (Fig.2).

Based on the characteristics of the utilized codes summarized in Table 3, it is normally expected
that ABAQUS and OPENSEES will give more or less similar results as both of them use implicit /
FEM numerical scheme, full Rayleigh damping and Masing rules. NL-DYAS on the other hand uses
quite different assumptions compared to the other three codes, as it uses an explicit finite difference
solver. Finally, in DEEPSOIL and NL-DYAS a constant viscous damping and no Masing rules are
utilized.
Table 2. Main characteristic of the utilized codes
Code

ABAQUS

OPENSEES

DEEPSOIL

NL-DYAS

Nonlinear model

Armstrong and Frederick

(1966).
Lemaitre and Chaboche
(1990)

Mazzoni et al. (2009)

2009

Hashash and Park (2001,

2002)

Gerolymos and Gazetas

(2005), Drosos et al. (2012)

Constitutive
Model
Kinematic
hardening model
combined with a
Von
Mises
failure criterion
and an associated
plastic flow rule
Anastasopoulos
et al, (2011)

Multi-yield
surface plasticity
(Ragheb, 1994;
Parra,
1996;
Yang, 2000)
MRDF pressuredependent
hyperbolic model
Hashash et al.
(2012)
Smooth
hysteresis model
originally
proposed
by
Bouc (1971) and
Wen
(1976).
Extended
by
Gerolymos and
Gazetas (2005)

Numerical
Scheme

Rayleigh
Damping

Masing
rules

Full

Yes

Full

Yes

Newmark
(1959)

method

No
(frequency
independe
nt
damping)

No

Explicit
Finite
Difference
Solver

No
(viscous
damping)

No

Implicit
FEM

Implicit
FEM

ELASTIC SOIL RESPONSE

Representative results of the soil response under purely elastic and visco-elastic behaviour using the
four numerical models are presented and discussed in this section. Fig.4 shows typical acceleration
time histories at the ground surface considering purely elastic response of the homogeneous soil
profile P1. The viscous damping is about 1% and the bedrock is simulated as elastic half-space using
appropriate dashpot values. Fig.4a illustrates the whole time history (15s-80s), while, to better
illustrate the differences, in Fig.4b a time window of the time-histories is presented referring to the
maximum computed acceleration (34s-36s). Fig.5a compares the computed horizontal accelerations
(time window from 34s to 36s), for profile P1 over elastic half-space assuming visco-elastic soil
response and the real accelerogram as input. Similar comparison is made in Fig.5b where the surface

A. Karatzetzou, S. Fotopoulou, E. Riga, S. Karapetrou, G. Tsinidis, E. Garini, K. Pitilakis, N. Gerolymos and G.

Gazetas
7

horizontal acceleration is presented for the soil profile subjected to a pulse signal and rigid bedrock
conditions.
It is observed that for both purely elastic and visco-elastic analyses all implemented numerical
codes predict quite similar responses. Numerical results are in good agreement with each other both in
terms of acceleration amplitudes and frequencies. Even for the low damping value (Fig.4) results in
terms of acceleration amplitudes almost coincide between the different codes.
Some numerical problems were detected in case of almost purely elastic analyses when rigid
bedrock is assumed. The unrealistically small damping considered in this case, resulted in small
attenuation of the computed acceleration after main shaking, in some codes. The problem was less
evident in the case of the elastic base, as part of the seismic energy was absorbed by the dashpots at
the model base.

(a)

(b)

Figure 4. Acceleration time histories at the ground level for elastic response of soil profile P1, elastic case for
damping, real record as input and elastic half-space (a) 15s-80s, (b) 34s-35s

(a)

(b)

Figure 5. Acceleration time histories at the ground level for profile P1 elastic soil response, visco-elastic case for
damping, (a) real record as input and elastic half-space and (b) pulse record as input and rigid bedrock

The soil resonant frequencies, derived from transfer functions, were found in good agreement
with the theoretical closed form solutions. More specifically, according to these solutions resonant
frequencies are estimated equal to 3.75 Hz and 1.16 Hz for profile P1 and profile P2, respectively.
Numerical analyses reveal similar resonant frequencies (Fig.6). For profile P1 (Fig.6a) all numerical
codes give exactly the same first frequency and amplitude. The second frequency is, however, slightly
lower when ABAQUS, OPENSEES and DEEPSOIL are used. This difference is however more
noticeable for ABAQUS and OPENSEES, which give about 5% lower value compared to the
corresponding theoretical expected frequency (11.25Hz). The second soil frequency is the same as the
theoretical one when NL-DYAS code is utilized, even if the amplitude for this frequency is up to 50%
lower compared with the other three codes. This can be attributed to the different approximations of
the codes concerning the viscous damping simulation.
In case of soil profile P2 (Fig.6b), all codes give almost similar results both in terms of
frequency and amplitude at least for the first three resonant frequencies. For higher resonant

frequencies (f > 6.5Hz) results differ mainly in terms of amplitude but also for frequencies. Again, this
should be expected due to the different damping formulation each code uses.
It is important to mention that for the elastic case and for the frequency range of interest in
engineering practice (f<6.5Hz) all studied codes give similar results.

(b)

(a)

Figure 6. Transfer functions (ground level to bedrock level Fourier) for elastic response of (a) P1 profile, viscoelastic case for damping, real record as input and elastic half-space d) P2 profile, visco-elastic case for damping,
real record as input and elastic half-space

Fig.7 depicts the peak horizontal acceleration with depth for profile P1 (Fig.7a) and P2 (Fig.7b).
Results refer to visco-elastic analysis using the real ground motion record assuming an elastic base. In
case of soil profile P1 results are in good agreement for the first 6 m except for NL-DYAS where the
acceleration amplitudes are smaller compared to the three other codes for a depth of 2m from the
surface. For depths greater than 6m the differences between ABAQUS, OPENSEES and DEEPSOIL
are still minor, while on the other hand NL-DYAS gives smaller accelerations values. This observation
is expected considering that in higher frequency values the amplitudes using NL-DYAS were lower
(see Fig. 6a); it is also due to the fact that at higher depths the input frequency range is higher. In case
of soil profile P2 (Fig.7b), all numerical codes predict in general comparable results with the
deviations between the results being in the order of 10% to 30%. At the ground surface of the 20m soil
profile P1 differences are not important (Fig.7a), while in case of the gradient profile P2 computed
accelerations differ over 35%. In particular, ABAQUS and OPENSEES compute lower acceleration
values at ground surface compared to DEEPSOIL and NL-DYAS. This observation may be attributed
to the different simulation of soil damping. Indeed, in case of deeper soil profiles Rayleigh damping
calibration affects more the computed response.
It is also noticed that ABAQUS and OPENSEES follow the same trend regarding the variation
of peak acceleration amplitudes with depth not only for profile P1 but also for P2 where the
differences between the various codes are more pronounced. At some depths NL-DYAS seems to
follow a slightly different trend compared to the other codes. Considering however the different
approximations of each code regarding the viscous damping simulation, the comparison of the peak
acceleration values with depth is considered overall satisfactory.
Representative maximum shear strain variation with depth is presented in Fig. 8. Results refer to
visco-elastic soil response of soil profiles P1 (Fig.8a) and P2 (Fig.8b), excited with the real ground
motion record. Shear strains are almost identical for both profiles with differences being on average 315% in all codes.

NONLINEAR SOIL RESPONSE

Fig. 9 presents comparisons of acceleration time histories for soil profile P1 computed at the ground
surface when subjected to the real record scaled at peak ground acceleration equal to 5 m/s2. To study
the effect of bedrock modeling (e.g. rigid or elastic bedrock), numerical results, referring to both rigid
8

A. Karatzetzou, S. Fotopoulou, E. Riga, S. Karapetrou, G. Tsinidis, E. Garini, K. Pitilakis, N. Gerolymos and G.

Gazetas
9

and elastic half space, are presented. Bedrock simulations have negligible effect on the frequency
characteristics of the computed time histories at soil surface. Actually, signals phase is not modified
significantly between the two cases, indicating that frequency characteristics are similar irrespectively
of the bedrock type. On the other hand, bedrock type seems to have an important effect on the
acceleration amplitude. Indeed, for the case of rigid bedrock, acceleration amplitudes are higher, as
theoretically expected.

(a)

(b)

Figure 7. Peak horizontal accelerations with depth for the elastic soil response, visco-elastic case for damping,
real record as input and elastic half-space for profiles (a) P1 and (b) P2

(a)

(b)

Figure 8. Maximum strain % with depth for the elastic soil response, visco-elastic case for damping, real record
as input and elastic half-space for profiles (a) P1 and (b) P2

Comparing different codes it is observed that ABAQUS, OPENSEES and DEEPSOIL result in
quite similar response in terms of frequency characteristics (similar phase). On the contrary NL-DYAS
results are quite distinct compared to the other codes. This observation indicates that stiffness is more
or less similarly captured by ABAQUS, OPENSEES and DEEPSOIL contrary to NL-DYAS.
As far as the amplitude is concerned, OPENSEES gives lower values. This implies that the
hysteretic damping is over-predicted by the model compared to the other codes, where the amplitudes
are similar. Although OPENSEES and ABAQUS are using similar constitutive models the calibration
procedures they use (automatic in OPENSEES and manual in ABAQUS), may be responsible for the
observed differences at the final computed response.
In case of soil profile P2 (Fig.10) the amplitudes of peak horizontal accelerations are again in
relatively good agreement between the different methods. However, the phases are quite distinct
indicating discrepancies in estimated stiffness of the soil deposits in different numerical codes.
Considering the high amplitude of the seismic input motion (0.5g at 100 m depth), the soil gradient
model and the way that this is modeled in each code, and differences on the way the non-linear soil
response is accounted for in each code, we may conclude that these rather limited differences should
be expected and they are in general within the engineering accuracy margin.

(a)

(b)

Figure 9. Acceleration time history (peak acceleration windows) at the ground level for profile P1 and non-linear
soil response for (a) elastic half-space and (b) rigid bedrock

(a)

(b)

Figure 10. Acceleration time history windows at the ground level for profile P2 non-linear soil response and (a)
elastic half-space and (b) rigid bedrock

Further conclusions can be drawn from Fig.11 where the maximum shear stresses, shear strains
and accelerations distributions along with depth are presented. Results concern profile P1 on a elastic
10

A. Karatzetzou, S. Fotopoulou, E. Riga, S. Karapetrou, G. Tsinidis, E. Garini, K. Pitilakis, N. Gerolymos and G.

Gazetas
11

half space. Considering the different hypotheses used by each code to account for soil non-linearity,
the results are again satisfactory. More specifically, the maximum differences are not exceeding 20%
for stresses at depth of 19m while this difference is further decreased towards the surface. In terms of
strains the results are generally in good agreement and the maximum difference, again at 19m, dont
exceed 50%. It is noticed however that this large difference concerns only this depth and only
DEEPSOIL and thus the results are considered satisfactory. Concerning the peak acceleration
amplitudes, the differences are rather important among the four codes; OPENSEES give the lower
values and NL-DYAL the highest at almost all depths (almost double compared to OPENSEES).
ABAQUS and DEEPSOIL give comparable results in all cases.

(a)

(b)

(c)

Figure 11. (a) Maximum stress (b) maximum stain (%) and (c) peak horizontal acceleration with depth for the
non-linear soil response of profile P1

CONCLUSIONS
The paper studied the elastic, visco-elastic and non-linear response of two idealized soil profiles: a
homogeneous single layer soil deposit of 20 m depth with constant shear wave velocity and a soil
profile of 100 m depth with gradient distribution of shear wave velocity with depth. The profiles were
subjected to a number of idealized (Ricker type) and really recorded time-history accelerations. The
numerical simulations were performed with NL-DYAS, DEEPSOIL, OPENSEES and ABAQUS,
incorporating several constitutive models. Results are presented in terms of transfer functions, peak
ground accelerations, shear stresses and shear strains.
For both purely elastic and visco-elastic analyses all four codes predict quite similar responses
in terms of peak acceleration amplitudes and frequencies even for low damping value. Some
numerical problems were detected only for the case of low damping value and rigid bedrock. In
frequency domain and especially in the frequency of interest in engineering practice (f<6.5Hz) all
codes give similar results. In higher frequencies ABAQUS, OPENSEES and DEEPSOIL slightly
underestimate the soils resonant frequency or the amplitude value.
The comparison for the non-linear ground response analysis is not always equally satisfactory;
the reason should be attributed to the differences in constitutive model and damping modeling between
the codes. The most important conclusions of the present work are summarized as follows: (i) Bedrock
simulation (e.g. rigid or elastic half space) have a negligible effect on the frequency characteristics of

the computed time histories at soil surface; on the contrary they affect the peak ground acceleration
amplitudes, with the rigid rock presenting clearly higher values. (ii) Comparisons of predicted
response indicate that (a) in terms of frequency content ABAQUS, OPENSEES and DEEPSOIL result
in quite similar results (similar phase); (b) in terms of peak ground accelerations ABAQUS and
DEEPSOIL predict similar values while OPENSEES and NL-DYAS give the lowest and highest
values in almost all depths; (c) all four codes predict quite similar maximum shear stresses and strains
values.
The use of different numerical codes to estimate site effects may lead to considerable
differences, which depend on several parameters not always related to the constitutive relationships
used in each code to model the non-linear soil behavior. For example rigid bedrock should be avoided.
The user should be aware of the limitations of each code and apply them consciously and according to
the problem anticipated.

REFERENCES
ABAQUS (2012) ABAQUS: theory and analysis users manual version 6.12, Providence, RI, USA: Dassault
Systmes SIMULIA Corp
Anastasopoulos I, Gelagoti F, Kourkoulis R, Gazetas G (2011) Simplified constitutive model for simulation of
cyclic response of shallow foundations: validation against laboratory tests. J Geotech Geoenviron Eng
137(12):11541168
Armstrong, P. J., and Frederick, C. O. (1966). A mathematical representation of the multiaxial bauschinger
effect. CEGB Rep. No. RD/B/N731
Bouc R. (1971). "Modle mathmatique d'hystrsis: application aux systmes un degr de libert". Acustica
(in French) 24, pp. 1625.
Drosos V. A., Gerolymos ., and Gazetas G (2012). ''Constitutive model for soil amplification of ground
shaking: Parameter calibration, comparisons, validation''. Soil Dynamics and Earthquake Engineering,
Vol.42, pp. 255-274.
Gerolymos ., and Gazetas G (2005). ''Constitutive model for 1-D cyclic soil behaviour applied to seismic
analysis of layered deposits''. Soils and foundations, Japanese Geotechnical Society, Vol.45, Is.3, pp. 147159.
Hashash YMA and Park D (2001) Non-linear one-dimensional seismic ground motion propagation in the
Mississippi embayment Engineering Geology, 62(1-3):185-206.
Hashash YMA and Park D (2002) Viscous damping formulation and high frequency motion propagation in
nonlinear site response analysis Soil Dynamics and Earthquake Engineering, 22(7): 611624
Hashash YMA, Groholski DR, Phillips CA, Park D, Musgrove M (2012) DEEPSOIL 5.1, User Manual and
Tutorial. 107 p.
Lemaitre, J., and Chaboche, J.-L.(1990). Mechanics of solid materials, Cambridge Univ. Press, Cambridge, UK
Lysmer J., A.M. Kuhlemeyer, Finite dynamic model for infinite media. Journal of the Engineering Mechanics
Division, ASCE, 95, 859-877, 1969.
Mazzoni S, McKenna F, Scott, MH, Fenves GL (2009) Open System for Earthquake Engineering Simulation
User Command-Language Manual, Pacific Earthquake Engineering Research Center, Berkeley,
California
Newmark NM (1959) A method of computation for structural dynamics ASCE, Journal of Engineering
Mechanics Division, 85:67-94
Park D and Hashash YMA (2004) Soil damping formulation in nonlinear time domain site response analysis,
Journal of Earthquake Engineering, 8(2): 249-274
Parra E (1996) Numerical Modeling of Liquefaction and Lateral Ground Deformation Including Cyclic Mobility
and Dilation Response in Soil Systems, Ph.D. Thesis, Dept. of Civil Engineering, Rensselaer Polytechnic
Institute, Troy, NY.
Ragheb AM (1994) Numerical analysis of seismically induced deformations in saturated granular soil strata,
PhD Dissertation, Dept. Of Civil Eng., Rensselaer Polytechnic Institute, Troy, NY
Schnabel PB, Lysmer J and Seed H.B (1972) SHAKE: A computer program for earthquake response analysis of
horizontally layered sites. Report no. UCB/EERC-72/12, Earthquake Engineering Research Center,
University of California, Berkeley.
Wen Y. K. (1976). "Method for random vibration of hysteretic systems". Journal of Engineering Mechanics,
ASCE, 102 (2), pp. 249263
Yang Z. (2000) Numerical Modeling of Earthquake Site Response Including Dilation and Liquefaction, Ph.D.
Thesis, Dept. of Civil Engineering and Engineering Mechanics, Columbia University, NY, New York.
12